Response from M-Systems

I have received another mail from Amir Ban at M-Systems, explaining their concerns with regard to releasing a GPL'd driver for the Disk-On-Chip hardware and for their NFTL flash filing system. It is unclear from this whether it is a final statement of their response, or whether they are still open to discussion.

From: Amir Ban <amirban@m-sys.com>
To: David Woodhouse <David.Woodhouse@mvhi.com>
Subject: RE: FW: Linux drivers vs. free software
Date: Fri, 9 Jul 1999 01:49:59 +0200

David,

I understand from this that you need the programming spec of the DOC2000.
Let me check with marketing if we can provide you with such a spec.

About the rest of it:

In general we are open and helpful to people who want to write code for the
DiskOnChip. If we GPL the driver source code, the concern is not that it may
be copied to work with the DiskOnChip, but that it may be used to work with
competing devices. The code contains, for example, a complete NAND Flash
MTD, NFTL, and a tested framework for handling a Flash block device complete
with solutions to several problems that our experience shows need a
solution, and a novice competitor may not even be aware of. While most of
the value lies in the hardware, we think the software represents significant
value even if it's not used to manage the DiskOnChip but something else.

If we release the software with no limitation, we can expect to see it soon
being used to drive us out of business. Low-tech Taiwanese companies will
construct a basic flash array and use our code to support it. Flash vendors
like Samsung and Intel stand to gain the most from this. They will likely
put the code on their web site and offer incentives to developers who can
apply to their Flash. Our patent for NFTL becomes irrelevant, since by
releasing it under the GPL we effectively waive our rights.

We may still have a business going on the strength of the hardware, but it
will be a smaller and tougher one.

Amir

• NFTL is patented. The existence of a GPL'd version of the code does not affect the patent. For a competitor to use NFTL in their devices would be verging on suicidal.
• Likewise, for Intel to put up a web page containing code, and actively encourage manufacturers to violate M-Systems' patents, would also be insane.
• The GPL is far from being "no limitation" - GPL'd code could not be reused in competitors' proprietary devices, unless they made the code available; in which case it'd be immediately obvious that they're violating the patent. And if they didn't publish the code, then M-Systems' lawyers would have to pick up the pieces that the Open Source people leave behind before they could sue them.
• Also, there has been a lot of public interest in this subject over the last few days - far more than I expected. If M-Systems don't release a basic driver for NFTL, then I expect it to be reverse-engineered fairly shortly. Any reverse-engineered version would be likely to go under the MIT license, allowing it to be freely used in competitors' proprietary code (patent issues aside). So by releasing a GPL'd version of NFTL, M-Systems would actually be restricting the distribution of it.

Possible solution

I think that that a possible way for us to proceed would be for M-Systems to provide technical specifications, rather than code. The basic details of how to access the flash on the Disk-On-Chip 2000 hardware, and the on-flash format of NFTL. We can reimplement the code, and any problems that we encounter on the way are entirely up to us, as are the clever parts of the wear-levelling algorithms and such like.

That way, we get a functional device under Linux, and M-Systems get to retain as much of their clever solutions as possible. Basically, all they're releasing is the stuff that would get reverse-engineered anyway. They win, because it's under GPL instead of under a BSD license, and we win, because we don't have to spend months beating on the hardware to reverse-engineer it.